Marios Thoma is a Research Associate in the Cognitive Artificial Intelligence and Robotics (CAIR) Research Group at the CYENS Centre of Excellence. He specializes in neural-symbolic AI and the development and behavioral evaluation of agentic pipelines based on large language models (LLMs), including tool-using agents and red-teaming protocols, with emphasis on LLM behavior characterization and failure-mode discovery. He also applies quality-diversity methods (e.g., MAP-Elites) to financial portfolio optimization. At CYENS, his projects have included PEDESTRIAN, developing an egocentric-vision dataset focusing on improving pedestrian safety, and several cross-platform smartphone apps using custom-trained computer-vision models for artwork identification and pedestrian-flow obstacle detection. Previously, at the Open University of Cyprus (Computational Cognition Lab), he worked on the MARI-Sense project, focusing on expert systems with neural-symbolic methods.